M3
86 records across 19 models (quants: 1bit, 2bit, 4bit, 8bit, F16, Q4_0, Q8_0, mxfp4, oQ4e, oQ4e-fp16).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| M3 | Bonsai-27B | 1bit | oMLX | 12.4 | 39.2 | – | – | 1024 | 2026-10-11 | community | – | source |
| M3 | Bonsai-27B | 1bit | oMLX | 12.1 | 38.2 | – | – | 1024 | 2026-10-11 | community | – | source |
| M3 | Bonsai-27B | 1bit | oMLX | 11.7 | 38.4 | – | – | 1024 | 2026-10-11 | community | – | source |
| M3 | Gemma 4 E2B | 4bit | oMLX | 61.5 | 1436.0 | – | – | 1024 | 2026-10-13 | community | – | source |
| M3 | Gemma 4 E2B | 4bit | oMLX | 58.1 | 1900.0 | – | – | 4096 | 2026-10-13 | community | – | source |
| M3 | Gemma 4 E4B | 4bit | oMLX | 20.2 | 527.3 | – | – | 1024 | 2026-10-13 | community | – | source |
| M3 | Gemma 4 E4B | 4bit | oMLX | 19.5 | 546.2 | – | – | 4096 | 2026-10-13 | community | – | source |
| M3 | Gemma 4 E4B | 8bit | oMLX | 15.6 | 519.1 | – | – | 1024 | 2026-10-10 | community | – | source |
| M3 | Gemma 4 E4B | 8bit | oMLX | 15.4 | 542.5 | – | – | 4096 | 2026-10-10 | community | – | source |
| M3 | gemma-4-26B-A4B-it-OptiQ-4bit-REAP-14B | 4bit | oMLX | 22.2 | 343.1 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 | gemma-4-E4B-it | 8bit | oMLX | 16.2 | 530.6 | – | – | 1024 | 2026-10-10 | community | – | source |
| M3 | gemma-4-E4B-it | 8bit | oMLX | 15.6 | 535.1 | – | – | 4096 | 2026-10-10 | community | – | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 35.6 | 551.3 | – | – | 1024 | 2026-10-10 | community | – | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 32.4 | 551.4 | – | – | 1024 | 2026-10-10 | community | – | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 31.1 | 557.9 | – | – | 4096 | 2026-10-10 | community | – | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 30.3 | 555.7 | – | – | 4096 | 2026-10-10 | community | – | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 16.9 | 539.3 | – | – | 1024 | 2026-10-10 | community | contradiction | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 16.7 | 539.8 | – | – | 4096 | 2026-10-10 | community | contradiction | source |
| M3 | gemma-4-E4B-it-qat | 8bit | oMLX | 15.5 | 479.6 | – | – | 32768 | 2026-10-10 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 52.3 | 503.1 | – | – | 1024 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 52.3 | 503.1 | – | – | 1024 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 51.7 | 502.2 | – | – | 1024 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 50.8 | 536.6 | – | – | 4096 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 50.8 | 536.6 | – | – | 4096 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 50.0 | 535.4 | – | – | 4096 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 48.1 | 510.6 | – | – | 8192 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 48.1 | 510.6 | – | – | 8192 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 42.7 | 498.7 | – | – | 4096 | 2026-10-16 | community | contradiction | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 41.9 | 447.1 | – | – | 16384 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 41.9 | 447.1 | – | – | 16384 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 24.8 | 255.1 | – | – | 32768 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 24.8 | 255.1 | – | – | 32768 | 2026-10-14 | community | – | source |
| M3 | LFM2.5-24B-A2B | 4bit | oMLX | 12.6 | 264.5 | – | – | 32768 | 2026-10-16 | community | contradiction | source |
| M3 | LFM2.5-8B-A1B | 4bit | oMLX | 68.7 | 805.5 | – | – | 4096 | 2026-10-16 | community | – | source |
| M3 | LFM2.5-8B-A1B | 4bit | oMLX | 51.0 | 617.5 | – | – | 32768 | 2026-10-16 | community | – | source |
| M3 | Ling-3.0-tiny | oQ4e | oMLX | 63.9 | 873.6 | – | – | 16384 | 2026-10-08 | community | – | source |
| M3 | Ling-3.0-tiny | oQ4e | oMLX | 48.1 | 690.7 | – | – | 32768 | 2026-10-09 | community | – | source |
| M3 | Ling-3.0-tiny | oQ4e | oMLX | 47.1 | 670.2 | – | – | 32768 | 2026-10-09 | community | – | source |
| M3 | llama 7B | Q4_0 | llama.cpp (ROCm) | 21.47 | 310.95 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | Q4_0 | llama.cpp (CUDA) | 73.38 | 1816.45 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | Q4_0 | llama.cpp (CUDA) | 69.83 | 1956.45 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | Q4_0 | llama.cpp (CUDA) | 69.49 | 1087.59 | – | – | tg128 | 2024-01-11 | community | – | source |
| M3 | llama 7B mostly | Q8_0 | llama.cpp (CUDA) | 50.75 | 1937.77 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | Q8_0 | llama.cpp (CUDA) | 50.47 | 1049.96 | – | – | tg128 | 2024-01-11 | community | – | source |
| M3 | llama 7B mostly | Q8_0 | llama.cpp (CUDA) | 42.43 | 1688.37 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | F16 | llama.cpp (CUDA) | 35.42 | 925.62 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | F16 | llama.cpp (CUDA) | 35.28 | 962.94 | – | – | tg128 | 2024-01-11 | community | – | source |
| M3 | llama 7B mostly | Q4_0 | llama.cpp (Metal) | 30.36 | 338.37 | – | – | tg128 | 2023-12-01 | community | – | source |
| M3 | llama 7B mostly | F16 | llama.cpp (CUDA) | 23.93 | 3022.37 | – | – | tg128 | 2024-01-21 | community | – | source |
| M3 | llama 7B mostly | Q4_0 | llama.cpp (Metal) | 21.34 | 186.75 | – | – | tg128 | 2023-11-27 | community | – | source |
| M3 | llama 7B mostly | Q4_0 | llama.cpp (CUDA) | 19.23 | 671.1 | – | – | tg128 | 2024-01-11 | community | contradiction outlier | source |
| M3 | llama 7B mostly | Q8_0 | llama.cpp (CUDA) | 19.11 | 637.51 | – | – | tg128 | 2024-01-11 | community | contradiction | source |
| M3 | llama 7B mostly | F16 | llama.cpp (CUDA) | 18.46 | 648.07 | – | – | tg128 | 2024-01-11 | community | contradiction | source |
| M3 | llama 7B mostly | Q8_0 | llama.cpp (Metal) | 16.46 | 339.31 | – | – | tg128 | 2023-12-01 | community | – | source |
| M3 | llama 7B mostly | Q8_0 | llama.cpp (Metal) | 12.27 | 187.52 | – | – | tg128 | 2023-11-27 | community | – | source |
| M3 | Mellum2.1-12B-A2.5B-Thinking | mxfp4 | oMLX | 57.3 | 554.7 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 | Ornith-1.5-9B | 4bit | oMLX | 18.8 | 194.7 | – | – | 1024 | 2026-10-08 | community | – | source |
| M3 | Ornith-1.5-9B | 4bit | oMLX | 18.7 | 202.5 | – | – | 4096 | 2026-10-08 | community | – | source |
| M3 | Ornith-1.5-9B | 4bit | oMLX | 18.1 | 195.0 | – | – | 8192 | 2026-10-08 | community | – | source |
| M3 | Ornith-1.5-9B | 4bit | oMLX | 15.4 | 177.6 | – | – | 16384 | 2026-10-08 | community | – | source |
| M3 | Qwen3.5-4B | 4bit | oMLX | 32.7 | 312.5 | – | – | 1024 | 2026-10-16 | community | – | source |
| M3 | Qwen3.5-4B | 4bit | oMLX | 31.5 | 323.8 | – | – | 4096 | 2026-10-16 | community | – | source |
| M3 | Qwen3.5-4B | 4bit | oMLX | 30.8 | 334.5 | – | – | 1024 | 2026-03-09 | community | – | source |
| M3 | Qwen3.5-4B | 4bit | oMLX | 29.5 | 296.7 | – | – | 4096 | 2026-03-09 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 19.1 | 194.0 | – | – | 1024 | 2026-10-13 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 19.0 | 196.0 | – | – | 1024 | 2026-10-08 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 18.9 | 195.7 | – | – | 1024 | 2026-10-14 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 18.7 | 202.3 | – | – | 4096 | 2026-10-14 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 18.5 | 193.7 | – | – | 4096 | 2026-10-13 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 18.5 | 201.8 | – | – | 4096 | 2026-10-17 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 17.7 | 174.7 | – | – | 1024 | 2026-03-09 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 17.7 | 174.7 | – | – | 1024 | 2026-03-09 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 17.6 | 157.4 | – | – | 4096 | 2026-03-09 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 17.6 | 157.4 | – | – | 4096 | 2026-03-09 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 17.3 | 199.6 | – | – | 8192 | 2026-10-17 | community | – | source |
| M3 | Qwen3.5-9B | 4bit | oMLX | 17.0 | 195.7 | – | – | 4096 | 2026-10-08 | community | – | source |
| M3 | Qwen3.6-35B-A3B-OptiQ-4bit-REAP-19B | 4bit | oMLX | 28.4 | 386.9 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 | Qwen3.6-35B-A3B-OptiQ-4bit-REAP-19B | 4bit | oMLX | 27.6 | 394.3 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 | Qwen3.6-35B-A3B-OptiQ-4bit-REAP-19B | 4bit | oMLX | 27.5 | 418.2 | – | – | 4096 | 2026-10-12 | community | – | source |
| M3 | Qwen3.6-35B-A3B-OptiQ-4bit-REAP-19B | 4bit | oMLX | 26.8 | 428.5 | – | – | 4096 | 2026-10-12 | community | – | source |
| M3 | Qwen3.8-27B | 4bit | oMLX | 5.4 | 50.4 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 | Qwen3.8-27B | 4bit | oMLX | 5.2 | 51.9 | – | – | 4096 | 2026-10-12 | community | – | source |
| M3 | Spark-X2.5-4B-Abliterated-Uncensored | oQ4e-fp16 | oMLX | 11.4 | 152.1 | – | – | 4096 | 2026-10-06 | community | – | source |
| M3 | Spark-X2.5-4B-Abliterated-Uncensored | oQ4e-fp16 | oMLX | 4.7 | 105.3 | – | – | 16384 | 2026-10-06 | community | – | source |
| M3 | Ternary-Bonsai-2-27B | 2bit | oMLX | 7.4 | 46.6 | – | – | 1024 | 2026-10-12 | community | – | source |
| M3 | Ternary-Bonsai-2-27B | 2bit | oMLX | 7.4 | 46.8 | – | – | 1024 | 2026-10-12 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085